Detroit Metropolitan Area
I build ML systems that don't just work in notebooks — they run at national scale, handle millions of real-world variables, and hold up under production pressure. At Ericsson, I deployed a reinforcement learning model that optimized antenna tilt across 600,000+ cell towers in 100+ US markets, lifting signal quality by 36% — with safety guardrails I engineered to prevent network disruption during the model's exploration phase. I compressed a 6-week mission-critical ML pipeline migration to 7 business days by automating code updates, monitoring, and reporting across 60+ Linux servers. I also built a zero-cost geospatial threat detection system — ingesting daily NOAA storm data, mapping high-risk zones to tower locations at geohash precision — enabling emergency response 2 days earlier than the previous manual process. My background is deliberately cross-domain: MS in Data Science + MA in Applied Economics (both University of Michigan), BA in Linguistics (Yale). That combination means I can move fluently between model architecture, statistical rigor, and business impact — which is what separates ML engineers who ship from ones who prototype. Technical toolkit: Python, PyTorch, Hugging Face, Ray/RLlib, FastAPI, SQL, Bash, Kubernetes, Streamlit, Tableau. Strong foundation in causal inference (diff-in-diff, IV) and production MLOps. Currently exploring Senior ML Engineer and Applied Scientist opportunities where I can build high-reliability AI systems that operate at scale. Open to FAANG, fintech (Stripe, Plaid, Square), cloud/data platforms (Databricks, Snowflake), and healthcare AI. Let's connect if you're building something that needs to work the first time — at scale.
• Deployed production RL model optimizing antenna tilt across 600K+ cell towers in 100+ US markets — lifted signal quality 36% while engineering safety guardrails that prevented network degradation during live exploration, directly impacting service reliability for millions of subscribers. • Compressed a 6-week mission-critical ML pipeline migration to 7 business days (83% faster) by engineering automation scripts for code updates, server management, and monitoring across 60+ Linux servers — zero production incidents during cutover. • Built zero-budget geospatial threat detection pipeline — ingesting daily NOAA storm data, mapping risk zones to 600K+ cell tower locations at geohash precision via Streamlit heatmap — enabling infrastructure threat response 2 days earlier than the previous manual process. • Delivered FastAPI proof-of-concept for Mobile World Congress demonstration 1 month ahead of deadline — evaluated kNN and scipy curve-fitting algorithms, redirecting cross-functional team toward a more viable AI-based visualization approach with time to pivot. • Prevented wasted engineering investment by systematically evaluating XGBoost and alternative ML approaches for legacy hardware energy estimation — identified insufficient data signal and documented findings that halted further investment in an unviable direction. • Partnered with AI researchers, solution architects, and PMs to optimize and maintain gigabyte-scale telecom ML pipelines — improved model training throughput and reliability while ensuring production systems met SLAs and business KPIs across 60+ markets.